p

ho
ks
Early Efficacy Trials,

or
W
Biomarkers, & Diagnostic Tests

V
A D
AA

Tatiana M. Prowell, MD
Breast Cancer Scientific Liaison, FDA
18

Asst. Prof. of Oncology, Breast Cancer Program, Johns Hopkins

20

Tatiana.prowell@fda.hhs.gov
Twitter: @tmprowell
 Disclosures

p
ho
• I have no financial interests to disclose.

ks
• I will not discuss off-label use of unapproved

or
agents.

W
V
A D
AA
18
20

www.fda.gov
2
 Outline

p
ho
• Early Efficacy Trials

ks
– Large First-in-Human Trials/Seamless Development

or
– Single-arm trials

W
– Selected Endpoints

V
• Response rate
A D
• Patient reported outcomes
AA

• Biomarkers for Patient Selection

• Companion and Complementary Diagnostics
18
20

• Continued Biomarker Discovery

www.fda.gov 3
 Discrete-Phase Drug Development

p
ho
ks
• Oncology drug development has historically passed

or
through 3 discrete steps:

W
– Phase 1: MTD, DLTs, preliminary efficacy

V
– Phase 2: Efficacy assessment for “go/no-go”
D
– Phase 3: RCTs designed to provide adequate
A
AA
efficacy/safety data to support drug approval
• Distinct phases have become blurred both in theory
18

and in practice.
20

www.fda.gov 4
 What Has Driven the Change?

p
ho
• Scientific advances resulting in more effective drugs

ks
and early biomarker/companion diagnostic

or
development

W
• Greater focus on pathways than tissue/site of origin

V
• Desire for greater efficiency in drug development
D
– Avoid delays inherent in discrete phase development
A
AA
– Patient interest/demand
– Clinician/investigator/cancer center demand
18
20

www.fda.gov 5
 OHOP Experience

p
• Stated objectives/endpoints/eligibility criteria/informed

ho
consent consistent with usual phase 1 trials, but sample

ks
size/nature of data collected/actual goals are not!

or
• Nature of expansion cohorts in these trials

W
– Dose/schedule refinement
–

V
Variety of of tumor types

D
– Variety of molecularly-defined subsets
A
– Other drug combinations
AA
– Numerous
– Large
18

– Sample size open-ended in some cases

20

www.fda.gov 6
 Regulatory Implications

p
ho
• What are the implications of an entire drug development

ks
program occurring within a first-in-human protocol?
–

or
Level of IRB scrutiny of these trials
– Meetings between FDA & companies

W
– Oversight by relevant disease experts/division within OHOP
–

V
Size and quality of safety database
–

D
Adequacy of data to support global regulatory approvals
• Should seamless designs be reserved for drugs worthy of
A
AA
breakthrough therapy designation?
• What level of independent oversight is appropriate to protect
18

patient safety?
20

www.fda.gov 7
 Critical Questions

p
ho
• Is there a compelling rationale for including multiple expansion cohorts?

ks
• Do the stated objectives and endpoints reflect the true goals of the trial,
and is the sample size range consistent with these objectives?

or
• Is there an appropriate statistical analysis plan for all stated endpoints?

W
• Are the eligibility criteria appropriately tailored to the expansion cohorts?
• Is there a defined end to the trial, both in terms of efficacy and futility?

V
• Does the informed consent reflect the current knowledge of safety and
A D
efficacy of the investigational drug (and other agents in the same class)?
• Is there a system in place to communicate with all investigators in a timely
AA
fashion?
• If the trial may be used for drug approval:
18

• Is there an independent oversight committee?

• Has there been communication with global regulatory agencies?
20

www.fda.gov 8
 Need for Independent Oversight

p
ho
ks
• Independent oversight committee needed for trials of

or
sufficient size/score to support regulatory approval

W
– Ensure standard patient protections

V
– Provide scheduled “pauses” to clean, review, and respond to

D
the data observed thus far in development program
A
– Improve transparency & address potential for bias
AA

– Ensure appropriate statistical rigor

18
20

www.fda.gov 9
 Outline

p
ho
• Early Efficacy Trials

ks
– Large First-in-Human Trials/Seamless Development

or
– Single-arm trials

W
– Selected Endpoints

V
D
• Response rate
A
• Patient reported outcomes
AA
18
20

www.fda.gov 10
 Single Arm Trials

p
ho
• Single arm trials have formed the basis for
>50% of accelerated approvals of oncology

ks
drugs

or
• Advantages

W
– With objective response rate, all responses are

V
due to the drug
A D
– Fewer resources/less time to complete
– Appropriate in refractory populations
AA

– Response rates may be marginal in refractory

patients and time-to-event endpoints are
18

confounded
20

– Poor characterization of safety (drug vs. disease)

www.fda.gov 11
Challenges to the Old Paradigm

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov Slide courtesy, G. Blumenthal 12

 When Randomized Trials May Not
Be Needed or Feasible

p
ho
• New drug with compelling biological rationale

ks
in a biomarker-selected patient population

or
• New drug with unprecedented response rates

W
in a patient population with few treatment

V
options A D
AA
• Approved molecularly-targeted drug being
tested in a patients with a rare tumor type and
18

expression of the relevant biomarker

20

www.fda.gov 13
 Objective Response Rate

p
ho
• Isolates treatment effect from natural history

ks
• Lends itself well to use in single-arm trials

or
• Does not account for stable disease

W
• Poses challenges to interpret in certain settings:

V
– Bone metastases
A D
– Peritoneal carcinomatosis
AA
– Immuno-oncology agents
• Has historically not been viewed by FDA as direct
18

measure of clinical benefit but has frequently

20

served as the basis for accelerated approval

www.fda.gov 14
Nature and Extent of Response Matters

p
ho
ks
or
W
V
A D
AA
18
20

www.fda.gov Slide courtesy, G. Blumenthal 15

When Response Is Direct Clinical
Benefit

p
ho
Where are the tumors that are responding?

ks
Regular approval granted based on clinical response rate (and duration), the cosmetic
improvement and the high likelihood of tumor related symptomatic relief

or
W
V
A D
AA
18
20

Vismodegib Response Romidepsin Response

Von Hoff et al., NEJM, 2009; Piekarz et al., JCO, 2009; 27:
361: 1164-72 5410-5417
www.fda.gov 16 16
Slide courtesy, J. Beaver
 Durability/Persistence of Response:

p
Nivolumab/Ipilimumab in Metastatic Melanoma

ho
• Features of the response matter:

ks
– Objective response rate

or
W
– Durability of response
– Persistence of response after

V
D
treatment discontinuation
A – Depth of response
AA

– Symptomatic improvement
18
20

Postow et al. N Engl J Med 2015; 372: 2006-2017

www.fda.gov 17
 Depth of Response:
Nivolumab/Ipilimumab in Metastatic Melanoma

p
ho
ks
or
W
V
A D
AA
18
20

Postow et al. N Engl J Med 2015; 372: 2006-2017.

www.fda.gov 18
 Patient-Reporting of Outcomes

p
ho
• Clinicians underreport

ks
patient symptoms.

or
• Patient-reported symptoms

W
demonstrate better

V
correlation than clinician-

D
reported symptoms with A
disease status.
AA

• PROs directly assess clinical

benefit. Key is how to
18

measure!
20

Basch, NEJM 2010; 362: 865-869.

www.fda.gov 19
 PROs: The Devil in the Details

p
ho
• Many issues to tackle with PRO instruments in

ks
general

or
– Reliability (test-retest)?

W
– Content validity (developed with patient input)?

V
D
– Appropriate recall period?
A
– Appropriate language translations?
AA

– Ability to detect change over time in response to

18

an intervention?
20

– Clinically meaningful score changes?

www.fda.gov 20
 Hurdles for PROs in Oncology

p
ho
• PROs pose unique challenges in oncology

ks
– Open-label designs

or
W
– Single arm trials

V
– Major drug-related toxicities
A D
– A lot of missing data (not at random!)
AA

– Tendency to limit trials to excellent PS patients

18
20

www.fda.gov 21
 Ruxolitinib for Myelofibrosis:

p
A Case Study in Response & PROs

ho
ks
or
W
V
A D
AA
18
20

www.fda.gov 22
Ruxolitinib: Change in Total Symptom Score

p
ho
ks
or
W
V
A D
AA
18
20

www.fda.gov 23
 So What Accounts for the Success?

p
ho
• Enrolled patients who were symptomatic from their disease (i.e.

ks
something to improve)

or
• Engaged early with FDA to discuss trial endpoints and registration

W
strategy
• Developed a simple PRO tool with a limited number of items of

V
D
importance to patients
A
• Captured and submitted real-time PRO data electronically (not
AA
burdensome, minimal missing data)
• Assessed response rate AND change in the symptom score and
18

relationship of the two, which was plausible

20

• Aimed for superiority and had a real statistical analysis plan

www.fda.gov 24
The Role of Biomarkers in Clinical
Trial Design

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov 25 25
 Biomarker Approaches

p
ho
Investigational
Test all No stratification by

ks
subjects marker results

or
Standard of care

W
Investigational
Marker -

V
Standard of care

D
Test all Stratify by
subjects marker results A Marker +
Investigational
AA
Standard of care
18
20

Test all Marker -

Enroll only
subjects marker + subset Investigational
Tezak Z et al. Personalized Medicine. 2010;7(5): 517-530. Marker + Standard of care 26
 Biomarker Trials:
Principles for Use in Drug Development

p
ho
• Biomarker is the major pathophysiological driver of the

ks
disease

or
• Biomarker is the known molecular target of therapy

W
• Preliminary evidence of harm from early phase clinical studies

V
in patients without the biomarker
A D
• Preliminary evidence of lack of activity from early phase
AA
clinical studies in patients without the biomarker
• Preliminary evidence of modest benefit in an unselected
18

population, but the drug exhibits significant toxicity (i.e.

20

unfavorable risk: benefit ratio)

www.fda.gov 27
 Predictive Enrichment =
Feasibility + Ethics

p
ho
Prevalence of Biomarker Response in Marker-Negative Patients

ks
(% of Marker-Positive Response)

or
0% 50%

W
Sample Size Ratio Sample Size Ratio
25% 16 2.6

V
50% 4 1.8
75%
DA 1.8 1.3
AA
100% 1.0 1.0
18
20

www.fda.gov 28
 Predictive Enrichment =
Feasibility + Ethics

p
ho
Prevalence of Biomarker Response in Marker-Negative Patients

ks
(% of Marker-Positive Response)

or
0% 50%

W
Sample Size Ratio Sample Size Ratio
25% 16 2.6

V
50% 4 1.8
75% A D 1.8 1.3
AA
100% 1.0 1.0

• Predictive enrichment is critical to the feasibility and success of a trial when responders
18

constitute a small fraction of the population.

20

• Must consider both the prevalence of the biomarker and the likelihood of response in
biomarker-negative patients for rational, ethical trial design.

www.fda.gov 29
 Biomarker or Test?

p
ho
• Biomarker: Biological feature to be measured

ks
• Test: Means to perform the measurement

or
W
• Example

V
A D
AA
18
20

www.fda.gov 30
 Companion Diagnostic

p
ho
• A companion diagnostic is essential to safe

ks
and effective use of therapeutic product

or
• Companion diagnostic and drug approvals are

W
usually given contemporaneously

V
D
• Requires submission of clinical trial data
A
AA
• Guidance to industry is available
18
20

www.fda.gov 31
 Questions to Consider When
Designing a Biomarker Trial

p
ho
• An ideal biomarker trial will teach you about the IVD (in
vitro diagnostic), the drug, and their interaction.

ks
• However, ethics and resources may not always allow you to

or
learn all that you want.

W
• Which of the following do you hope to learn about the IVD?

V
– Sensitivity of IVD: what fraction of responders are biomarker

D
positive?
A
– Specificity of IVD: what fraction of non-responders are
AA
biomarker negative?
– PPV of IVD: what fraction of marker positive pts respond?
– NPV of IVD: what fraction of marker negative pts do not
18

respond?
20

www.fda.gov 32
 Questions to Consider When
Designing a Biomarker Trial

p
ho
• An ideal biomarker trial will teach you about the IVD (in vitro
diagnostic), the drug, and their interaction.

ks
• Ethical considerations and resource availability will impact trial

or
feasibility.
• Which of the following do you hope to learn about the IVD?

W
– Sensitivity of IVD: what fraction of responders are biomarker positive?

V
– Specificity of IVD: what fraction of non-responders are biomarker

D
negative?
A
– PPV of IVD: what fraction of marker positive pts respond?
AA
– NPV of IVD: what fraction of marker negative pts do not respond?
– Where should the cutoff be set?
• Trials limited to marker-positive patients characterize the
18

drug/diagnostic interaction poorly.

20

www.fda.gov 33
 Regulatory Considerations for
Companion Diagnostics

p
ho
• Approval of companion diagnostic is based upon clinical trial

ks
population.

or
• What type of claim is sought by the company for the diagnostic?

W
– Selection Claim: Diagnostic test can reliably select same population
enrolled in the trial

V
• Inclusion of marker-negative patients preferred, but not

D
required. A
– Predictive Claim: Presence of biomarker predicts response
AA

• Inclusion of marker-negative patients required!

• Ideally done in trials prior to approval, but may be done in post-
18

marketing setting
20

www.fda.gov 34
 Why We Should Include
Biomarker-Negative Patients

p
ho
• Even when we (believe that we) understand the science, we still get

ks
proven wrong. Often.

or
– May incorrectly assume biomarker is required for response deny

W
effective therapy to marker-negative patients
• Even if less effective in marker-negative, may still be meaningful

V
D
treatment effect in diseases with unmet need
A
– May select the wrong biomarker abandon development of a
AA
potentially active drug if no efficacy observed
– May be more than one biomarker predictive of response
18

• Understand consequences of potential off-label use

20

• To better characterize the diagnostic and understand diagnostic/drug

interaction
www.fda.gov 35
PALOMA-1 Trial Schema

p
ho
ks
or
W
V
D
A
AA
18
20

www.fda.gov 36
 PALOMA-1 Results:
PFS K-M Curve in ITT Population

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov 37
 PALOMA-1 Results:
PFS K-M Curve in Biomarker Positive

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov 38
 PALOMA-1 Results:
PFS K-M Curve in Biomarker Negative

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov 39
 Complementary Diagnostic

p
ho
• Less well-defined

ks
– Not essential for safe and effective use of the drug

or
– Generally used as stratification element for efficacy

W
in all-comers trial

V
D
– Often identified in post-approval discovery process
A
AA
18
20

www.fda.gov 40
 p
So your targeted agent got approved…

ho
ks
Can you (finally) stop thinking about

or
biomarkers?

W
V
A D
AA
18
20

www.fda.gov 41 41
 Case Study

p
ho
ks
or
W
Crizotinib
V
A D
AA
18
20

www.fda.gov 42
 Crizotinib Initial Approval

p
ho
• Crizotinib approved in 2011 under accelerated

ks
approval for the treatment of metastatic NSCLC that

or
is ALK-positive as detected by an FDA-approved test

W
• Approval based on ORR (with supportive DoR data)

V
– 2 single arm studies
– ORR 50% and 61%
A D
AA

• Post-marketing requirement included a confirmatory

18

control to establish clinical benefit

20

www.fda.gov 43 43
 Crizotinib Post-Marketing

p
• Confirmatory RCT of crizotinib vs. standard of care in

ho
2nd line ALK+ NSCLC with primary endpoint of PFS

ks
– Median PFS improved from 3 to 7.7 months (HR 0.49)

or
– Regular Approval granted in 2013 with benefit confirmed

W
in ALK+ NSCLC

V
• Approved for ROS1-positive NSCLC (breakthrough
therapy) in 2016
A D
AA

– 66% ORR in single arm 50 patient study

– Laboratory test for companion diagnostic is under
18

development
20

www.fda.gov 44 44
 Crizotinib in
ROS1-Rearranged NSCLC

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov A. Shaw et al, NEJM 2014 45

 Case Study

p
ho
ks
or
W
Trastuzumab
V
A D
AA
18
20

www.fda.gov 46
 This Is Not Necessarily A Fast Process…

p
ho
• Trastuzumab

ks
– 1979: HER2 oncogene first described (I was 7 years old)

or
– 1984: HER2 protein first discovered

W
– 1990: First animal studies of trastuzumab
–

V
1992: First in human trial of trastuzumab launched

D
– 1995: Randomized phase 3 MBC trial launched
–
A
1998: BLA filed for use in HER2+ MBC in May and FDA approval granted
AA
in September
– 2000: First adjuvant randomized phase 3 trial launched
18

– 2006: FDA approval in adjuvant setting

20

– And then…

www.fda.gov 47
NSABP B-31 K-M Curve

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov
E. Romond et al, NEJM 2005 48
 NSABP B-31 Efficacy by
Central HER2 Testing Results

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov L. Fehrenbacher et al, ASCO 2013 49

NCCTG 9831 K-M Curve in
HER2 Low Patients

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov L. Fehrenbacher et al, ASCO 2013 50

 NSABP B-47 Trial

p
ho
ks
or
W
V
AD
AA
18
20

www.fda.gov L. Fehrenbacher et al, ASCO 2013 51

 NSABP B-47 IDFS Results

p
ho
ks
or
W
V
AD
AA
18
20

L. Fehrenbacher et al., SABCS 2017

www.fda.gov 52
 NSABP B-47 OS Results

p
ho
ks
or
W
V
AD
AA
18
20

L. Fehrenbacher et al., SABCS 2017

www.fda.gov 53
 And So Almost 4 Decades Later…

p
• 1979: HER2 oncogene first described

ho
• 1984: HER2 protein first discovered

ks
• 1990: First animal studies of trastuzumab

or
• 1992: First in human trial of trastuzumab launched

W
• 1995: Randomized phase 3 MBC trial launched
• 1998: BLA filed for use in HER2+ MBC in May and FDA approval granted

V
D
in September
•
A
2000: First adjuvant randomized phase 3 trial launched
AA
• 2006: FDA approval in adjuvant setting
• 2011: NSABP B-47 trial of trastuzumab in HER2 low/normal launched
18

• 2017: NSABP B-47 shows no benefit in HER2 low/normal early BC

20

www.fda.gov 54
 Conclusions

p
• The trial design, endpoints, and development program

ho
should be tailored to the drug and pt

ks
population/indication sought.

or
• Start thinking about biomarkers early and keep an open

W
mind
– Not every drug will have an identifiable, testable biomarker

V
D
– Not every biomarker will be essential to safe and effective use
A
of the drug
AA
– Some drugs will have more than one biomarker
– The best biomarker may not be the first one you find
18

• You may never be finished thinking about any of this

20

(sorry…that’s what wine is for)

www.fda.gov 55
 Acknowledgements

p
ho
• Julia Beaver

ks
• Elizabeth Mansfield

or
W
• Raji Sridhara

V
A D
AA
18
20

www.fda.gov 56